• in silico models
• optimal cellular construct
• genetic algorithms

Advanced in vitro models have been developed for mimicking the functionality, the biomechanics and the architecture of human tissues. Despite they are widely considered a promising technology for different biomedical applications, quantitative design criteria for determining their physiological relevance and thus being considered functional replicates of their in-vivo counterpart are still lacking. Such physiologically relevance can emerge as the coherence to some biological and well-known biophysical rules. Specifically, it is important to ensure that all the cells within the vessel-free constructs have an adequate nutrient supply. Moreover, thermodynamical laws govern energy and mass distribution: indeed, an organism always reaches a configuration characterised by a minimal surface energy. These constraints must be considered in reproducing in-vitro experiments. In-silico methods intervene as cheap and fast methods through which a great amount of models could be studied in order to obtain results that help scientists in selecting the best experimental setup for in-vitro protocol.
In this scenario, a possible approach for defining the design criteria are the rules of ‘natural selection’, i.e., the process through which living organisms adapt and change. Mathematically speaking, it could be described as a global optimization process (GOP): iteratively, using exploitation and exploration mechanism, the global maximum or minimum point is found. One of the possible GOP is the so-called Evolutionary Algorithms (EA), where Genetic Algorithms (GA) are his sub-branch.
In this work, we investigated the feasibility of using GA as optimization tool for designing physiologically relevant in silico 3D cell-laden spheroids. Hence the identification of the optimal biophysical parameters for generating 3D in vitro constructs with an enhanced predictive and translational power. In addition, GA was enriched for evaluating how the optimal constructs’ shapes and sizes change when accounting also for environmental perturbations.
LiveLinkTM for Matlab, which extend the Comsol modelling environment with an interface between the two tools.
Briefly, a population of individuals is generated. Each individual is characterized by its own genome, whose genes are described as an array of morphology and biophysical constraint information. All the individuals were scored through the Fitness Function: highest the value, better the responding to the predefined biophysical constraint.
From this point, all the next generations are obtained through gene recombination application using crossover and mutation operators. The loop of evolution ends until maximum number of generation is reached, and thus the best score fitting construct is identified.
Having an in-silico protocol that automatically predicts optima shape and dimension of cellular construct, given some constraint describing physiological aspect, i.e., oxygen concentration inside the aggregate, will permits to drastically improve and speed up biomedical experimental protocol involving cells aggregates as principal character, like tissue engineering, drug and cosmetic testing for example. This will lead to reduction in terms of costs and time and will improve reproducibility and shareability of experiments in laboratories all over the word. Evolutionary algorithms approach is largely used nowadays in designing reconfigurable organism with the goal of optimize certain cellular behaviours under some constraints, providing powerful paradigm in connecting in-silico and in-vitro applications. In this work, a first attempt at exploiting GA for designing 3D cellular aggregate is presented. Preliminary results suggest the feasibility of this approach as first step for an in-silico trial to obtain the optimal design criteria for generating in-vitro physiologically relevant models.